Towards Automatic Recognition of Emotional States of Animals

Automatic recognition of human emotions is a vibrant multi-disciplinary topic attracting increasing interest in many scientific communities due to its numerous applications in healthcare, human-machine interaction, decision support and many more. Analogously to humans, animals experience many emotions, some of which are universal across species. Recognition of animal emotions is crucial for assessment of animal welfare and has many applications in veterinary science and applied behavior. Yet there is very little research on automatic recognition in this context. This research aims to take the first step in investigating and developing AI-based approaches for automatic recognition of animal emotions based on visual modalities: images and videos. Our main focus will be on emotions of positive/negative valence, and on pain. As such, the developed machine learning approaches have the potential for far reaching applications, e.g., for veterinary science (in the context of pain recognition) and automatic monitoring of welfare of sheltered animals.